April 1996 - Test - lists.nanog.org

Re: Some corrections
by William Allen Simpson 16 Apr '96

16 Apr '96

> From: Dorian Kim <dorian(a)cic.net> > On Mon, 15 Apr 1996, William Allen Simpson wrote: > > I am admittedly confused here. CICnet provides connectivity from UMich > > to OSU, and the other Big Ten schools. CICnet is physically connected > > to UMich and MichNet, and also to OSU and OARnet. > > This used to be the case. This is right now not the case. This maybe the case > again in the future. Confused yet? Welcome to the world of academic > networking. :) >... > Depends on what you mean. AS266 traffic goes through MCI to get to AS1225. >... > I'm still curious where you got this false impression of the situation. > Apparently, I was missing some pieces (that had not been described in your earlier messages), and remembering "old" information (as above). Both Merit and CICnet are connected to MCI. Correct? My new understanding is that the problem seems to be that MCI is stuck in the middle there, and for some reason shunts some (but not all) OSU and OARnet traffic off through MAE-East, rather than passing it more directly to OARnet (through CICnet or otherwise). And this is because of some _source_ based algorithm, since this only happens to Merit sourced traffic, but not other CICnet traffic, as you are using MCI to hold together pieces of CICnet? Sounds bizarre to me.... In any case, it still proves my point -- Merit and CICnet (and MCI) somehow view each other as competitors and don't cooperate to carry traffic by the "best" route, thereby clogging other areas of the Internet (MAE-East) at others' unknowing expense. Indeed, this sounds even closer to the UK situation, where they are too cheap to install a local European link, and instead steal from the other vendors' links to Europe through the MAE-East commons. WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

1 0

35th IETF Report Available
by lscanlan＠cix.org 16 Apr '96

16 Apr '96

An extensive summary report of the 35th IETF, held in March in Los Angeles, California, is now available at the CIX web site, http://www.cix.org/Meetings/ietf-35.html Highlights of the meeting topics covered by Mark McFadden, CIXtra's technical reporter, include: Explorations of Alternative Routing and Topology IRAD Portable Site Prefixes Exchanges Stratum-based Aggregation Globally Redundant Routing Information DNS-based Logical Routing Conclusion Internet Architecture Board Character Sets in the Internet IAB, IRTF and IETF Issues Hearing on Returning a Standard to Draft Standard Status Applications and Transport Joint Session "Netscape Hack" HTTPng Caching T/TCP Transaction TCP HTTP 1.0 Short-term Solutions HTTP Version 1.1 HTTP and Multicasting Other Work Ipng Working Group IP Security IPv6 Mobility Routing Aspects of the IPv4 to IPv6 Transition Agents, Mobile Code and Intelligent Agents CIDRD Lou Scanlan Member Information and Services Email: helpdesk(a)cix.org Voice: 703 8 CIX CIX

1 0

Re: Some corrections
by William Allen Simpson 15 Apr '96

15 Apr '96

> From: Dorian Kim <dorian(a)cic.net> > > Likewise, a lot of traffic from Ann Arbor Michigan to Columbus Ohio > > travels via MAE-East, despite the fact that Merit is already connected > > to CICnet, which is in turn connected to Columbus (both OSU and OARnet). > > The problem is, Merit has no "bi-lateral peering" with CICnet. > > This is not exactly correct. CICNet Michigan(AS266) peers with Merit over a > FDDI ring in Ann Arbor. > > I suspect that the problem you are running into is that CICNet Primary > region (AS1225) which does not have physical connectivity to AS266 or Merit > provides connectivity for OSU, and peers with OARnet. > I am admittedly confused here. CICnet provides connectivity from UMich to OSU, and the other Big Ten schools. CICnet is physically connected to UMich and MichNet, and also to OSU and OARnet. Are you saying that despite the peering with Merit/MichNet and the traffic actually flowing to OSU, some other pieces of CICnet at the same location (OSU and OARnet) are physically and topologically discontiguous? That is, you don't carry your own internal traffic? > This problems has nothing to do with who has what bilaterals with whom, or who > is willing to peer with whomever else, but rather a simple constraint based > on topology. > > So please choose a more apropos example next time. Thanks. > The quoted "bi-lateral peering" was taken from your own private message on the subject. Perhaps I don't understand the use of the term "bi-lateral peering". I thought that it meant willing to forward all traffic from each party to the other party. Your private message indicated that you have such an agreement with OARnet, but not with Merit, and that it was Merit that is unwilling to make the agreement. I then checked with Merit (John Vollbrecht), who gave me a similar (though not identical) explanation. Topological contraints were not mentioned in your private message. Sorry if I misunderstood your message. WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

2 1

Re: The Attitude (was: the Internet Backbone)
by Justin W. Newton 15 Apr '96

15 Apr '96

At 10:50 AM 4/14/96 -0700, Internet Architect wrote: >Bill, > > A short answer to your refusal to peer objection in SPRINT's >policies in the form of a question. Which is better: > >1. peering with everyone at an exchange, including those ISPs who are >clearly clueless, and whose cluelessness leads to operational problems, >i.e. injection of bogus routes, black holes, routing loops, routing >flaps, and BGP peer transitions? Clue and size have very little to do with each other. I setup and ran a router at MAE east for a few months and am about to setup routers at MAE East and Pennsauken. To date I have /never/ blown someone else's routes into someone else unless I was intentionally readvertising those routes as per someone's request. My router was properly dampened, didn't cause routing loops etc etc. That is /not/ the basis of Sprint's policy. Sprint's policyu is based on the I'm big and you're not philosophy. If peering was based on properly making announcements and not flapping etc etc, sprint themselves wouldn't have a lot of peers right now. (Sprint has in the past advertised routes that were not theirs, and were in fact mine on numerous occasions). As for stability, well, things have gotten a lot better. > >2. qualifying each of your peers as being clueful, prior to peering >with them, such that you (and your customers!) don't suffer from the >cluelessness of others? Sounds like a great idea. Justin Newton * You have to change just to stay Internet Architect * caught up. Erol's Internet Services *

1 0

The Attitude (was: the Internet Backbone)
by William Allen Simpson 15 Apr '96

15 Apr '96

Catching up on my email after tax week (did anyone else notice that there were a lot fewer postings on this and other lists this past week?): > From: Sean Doran <smd(a)icp.net> > It's because I'm an evil asshole determined to protect my > employer's interests and make our shareholders rich. > > This is somewhat incompatible with protecting the interests > of our competitors and enriching their shareholders. > Gentlefolk, while I agree that Sean has been "mostly" right on many technical issues, I am seriously unhappy with this attitude! Nor is this attitude unique to Sean. Since my local ISP and NSP is Merit/MichNet, and I am an advocate of more regional exchanges, I have been trying to get them to interconnect with other local providers, so that local traffic wouldn't have to go through MAE-East (and suffer 30%-50% losses) to travel 3 blocks across town. They never managed to do anything in 3 years, despite the willingness of others (specifically MSEN) to interconnect. Recently, I asked why they aren't connected to the Detroit NAP. The response was "everyone else should connect to MichNet, and pay our affiliate fee". I noted that the others consider themselves competitors, and taking the same attitude would expect Merit to instead pay THEM, since MichNet generates the most traffic. Likewise, a lot of traffic from Ann Arbor Michigan to Columbus Ohio travels via MAE-East, despite the fact that Merit is already connected to CICnet, which is in turn connected to Columbus (both OSU and OARnet). The problem is, Merit has no "bi-lateral peering" with CICnet. Merit doesn't think there is a "cost benefit" to have regional interconnection and peering relationships. In some respects, they are right. The benefit is not to Merit itself, but rather to its customers (lower delays), and the rest of the Internet (less congestion at other exchanges). The problem is that ISPs are allowed to shove their regional connectivity out to others on the Internet. In effect, the _rest_ of the Internet is _paying_ for the regional underprovisioning. > Personally, I have little patience for the small and > not-very-clueful who want to be direct competitors with a > multibillion dollar company with lots of talent and who are > taken to whining about my policies and those of my > colleagues and associates, and even those of our > competitors. This uncharitable attitude obviously does not > endear me to them. > But your attitude that the whiners are "small and not-very-clueful" is less than useless. There are some quite clueful folks that don't agree with your policies, particularly with the failure to peer (and exchange traffic) with everyone else (even small folks) at an exchange. (Sprint is not the only perpetrator of this poor policy.) The fact is, whether you like it or not, they _ARE_ your competitors in their specific regions. But, to thrive, the Internet has a long tradition of _cooperation_ among competitors. Kinda misses the meaning of "exchange". That hurts everyone else on the net, by increased delay and more congestion elsewhere. In short, you are asking _others_ to bear the costs of _your_ making money. We've seen this time and again, such as the UK provider who sends all their traffic to the US, which then uses the congested US to Europe links. It only saves them money because others were unknowingly bearing the cost. Sounds like a form of fraud to me. > I would hope, though, that the bulk of our customers > would be much happier with us driving towards a network > reliable enough that they don't have to worry about their > customers screaming (not to mention not having to worry > about facing some very difficult scaling problems we are > already staring at), than with us being the Department of > Warm and Fuzzy Feelings. > > All of them know full well that the drive ain't easy. > True. But there are some particular bones to pick with Sprint, like the underprovisioned Texas links that kept dropping out, just when Apple released its 7.5.3 MacOS Update to developers from Texas.... So, let's see some of that vaunted reliability first, please. WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

5 4

Re: More on 'Number of Tiers' Discussion
by Michael Dillon 14 Apr '96

14 Apr '96

On Sun, 14 Apr 1996, Stan Barber wrote: > I didn't type it in. I just cut and pasted it. Here it the text of the original > note you sent: > Have a look at http://www.sidhe.memra.com/rough.txt if you want to see Mea culpa. This should read http://sidhe.memra.com/rough.txt Michael Dillon Voice: +1-604-546-8022 Memra Software Inc. Fax: +1-604-546-3049 http://www.memra.com E-mail: michael(a)memra.com

1 0

More on 'Number of Tiers' Discussion
by Tim Bass 14 Apr '96

14 Apr '96

Food for Thought. Last week or so there was a suggestion here that 'we' adopt a four level tier model for hierarchical routing. I wanted to follow up, just a little, to explain to 'anyone interested' what is the generally accepted 'thinking' on hierarchical routing in large networks today. Kamoun and Kleinrock published an authoritative work on hierarchical routing and came to the interesting discovery: The optimal number of levels (tiers was used recently) is the natural log of the number of routed subnets, or just simply: ln(N) which reduces the routing table to the theoretical optimal number: e*ln(N) If we assume 'the four tier model' as suggested by a poster last week on NANOG, an upper bound is placed on the optimal number of routed subnets in a hierarchical model, basically: 4 = ln(N) or 54 subnets, in the K&K optimal case ;-) Which, BTW would be good for the routing table for an Inter-Hierarchy Routing Protocol (IHRP, for fun :-) .... for 148 entries ! If we, for fun, take the old class of IPv4 address space and simplify the numbers, without worrying about nits; 128 + (191-128)*2^8 + (255-193)*2^16 yields a ballpark number of 123+16128+4063232 ~= 4080 K Hence, ln(4080K) equals 15. So 15 is the optimal number of 'tiers' or levels in the current IPv4 address space (not taking into account distance-path vectors and other fun stuff). Now, of course, we can argue path-length trade offs vs. number-of-levels until our fingers are tired of typing e-mail, but why? The simple idea for illumination in this post is simply; It is not A Good Idea to propose a concrete, structured four-level superstructure for the Internet (and we have not begin to look at IPv6, BTW). This is all just theoretical mumbo-jumbo, BTW. Path length trade-offs, religion, the phase of the moon, greed, genetics, and every other conceivable human condition make ln(N) unobtainable. Our great^N grandchildren will be living in total peace and harmony before humankind will ever see ln(N). But, IMO, 'a four-tier model' is not What We Want, or WWW (note 1). Best Regards, Tim note 1: (a new image for WWW, the old one is getting stale, don't you think?) +------------------------------------------------------------------------+ | Tim Bass | | | Principal Network Systems Engineer | "Never underestimate the bandwidth| | The Silk Road Group, Ltd. | of a station wagon full of tapes | | | hurling down the highway." | | http://www.silkroad.com/ | -Andrew Tanenbaum | | | | +------------------------------------------------------------------------+

2 2

CSPAN at NANOG?
by William B. Norton 12 Apr '96

12 Apr '96

Hi all - Thanks for your feedback regarding CSPAN at the next NANOG. Based on the concerns that some of you raised I've decided *not* to send out the media advisory to CSPAN and the press. Issues leading to that decision included 1) attendees holding their tongues for fear of being misquoted, "Joe X from NetQ said '....' might be stated more succinctly as "NetQ representatives said '.....' or "NetQ said '....' 2) the press misunderstanding and then overemphasizing some issues There was also a fear that marketing folks might be attracted to the bright lights and degrade the technical quality of the meeting. Reasonable concerns. Cheers - Bill ---------------------------------------------------------------------------- ----- William Norton <wbn(a)merit.edu> (313) 936-2656

2 2

Re: CSPAN at NANOG?
by Paul Ferguson 12 Apr '96

12 Apr '96

Thank you. The voice of sanity prevails. :-) - paul At 08:38 AM 4/12/96 -0400, William B. Norton wrote: >Hi all - > >Thanks for your feedback regarding CSPAN at the next NANOG. Based on the >concerns that some of you raised I've decided *not* to send out the media >advisory to CSPAN and the press. Issues leading to that decision included >1) attendees holding their tongues for fear of being misquoted, > "Joe X from NetQ said '....' might be stated more succinctly as > "NetQ representatives said '.....' or "NetQ said '....' >2) the press misunderstanding and then overemphasizing some issues > >There was also a fear that marketing folks might be attracted to the bright >lights and degrade the technical quality of the meeting. Reasonable concerns. > >Cheers - > >Bill >---------------------------------------------------------------------------- >----- >William Norton <wbn(a)merit.edu> (313) 936-2656 > > >

1 0

Re: the Internet Backbone
by Sean Donelan 10 Apr '96

10 Apr '96

> People are out to get smd for the same reason they have been out > to get people throughout history. He's smart, he's right, and he's > moving the world forward. People aren't out to get smd. Most people don't even know or care who smd is. Customers know and care about the results they get. Not who got them, or how many billions was spent getting them. Naturally, its up to each network provider to decide how they want to run their network and their company. If they do an adequate job, their customers will be happy, and their employer and shareholders will be happy. Different providers make different choices running their network. Some providers think they run a better network sending traffic through a few circuitous routes. Other providers think they run a better network using more direct routes. History will tell which providers made the right choice. P.S. I wonder why Mr. Doran thought I was speaking of Sprint when I mentioned complexity for complexities sake. I was actually thinking about the ATM network fiascos. I'm glad to hear talented people at multi-billion dollar companies think only the complexity needed to get the job done was used by talented people at multi-multi-billion dollar companies. -- Sean Donelan, Data Research Associates, Inc, St. Louis, MO Affiliation given for identification not representation

1 0